NeurIPS 2020

Simplify and Robustify Negative Sampling for Implicit Collaborative Filtering


Meta Review

The initial reviews were mixed for this paper. However, during the discussion, a certain consensus emerged regarding the value of this contribution. In particular, the reviews agree that the proposed method is simple and effective. In the proposed study, the method seems to outperform (by a small margin) others consistently. Congratulations! The main negative is the high computational and memory cost of the approach. Reviewer #1's overall score seems to have been greatly influenced by this and his assessment seems correct. While we cannot expect all new methods to be better in all aspects, the reviewers found that a proper discussion of these aspects should be provided in the main paper (and not "relegated" to the appendix). I also strongly suggest that the authors consider the reviews to improve their submission. I found the post-rebuttal comments of Rev. #2 to be particularly relevant: "as the literature has two major experimental settings (using only 4+5 star ratings as positive or using all ratings a positive), it might be worthwhile to provide results for both in the final version of the paper (maybe in an appendix)."